Research Article
How Do You Learn to Walk? Thousands
of Steps and Dozens of Falls per Day
Psychological Science
23(11) 1387­–1394
© The Author(s) 2012
Reprints and permission:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/0956797612446346
http://pss.sagepub.com
Karen E. Adolph, Whitney G. Cole, Meghana Komati,
Jessie S. Garciaguirre, Daryaneh Badaly, Jesse M. Lingeman,
Gladys L. Y. Chan, and Rachel B. Sotsky
New York University
Abstract
A century of research on the development of walking has examined periodic gait over a straight, uniform path. The current
study provides the first corpus of natural infant locomotion derived from spontaneous activity during free play. Locomotor
experience was immense: Twelve- to 19-month-olds averaged 2,368 steps and 17 falls per hour. Novice walkers traveled
farther faster than expert crawlers, but had comparable fall rates, which suggests that increased efficiency without increased
cost motivates expert crawlers to transition to walking. After walking onset, natural locomotion improved dramatically:
Infants took more steps, traveled farther distances, and fell less. Walking was distributed in short bouts with variable
paths—frequently too short or irregular to qualify as periodic gait. Nonetheless, measures of periodic gait and of natural
locomotion were correlated, which indicates that better walkers spontaneously walk more and fall less. Immense amounts
of time-distributed, variable practice constitute the natural practice regimen for learning to walk.
Keywords
infant development, learning, motor processes, perceptual motor coordination
Received 9/5/08; Revision accepted 3/8/12
How do infants learn to walk? For more than 100 years,
researchers have described developmental antecedents of
walking, improvements in the kinematics of walking gait, and
changes in the neurophysiological correlates of walking
(Adolph & Robinson, in press). However, a century of research
has proceeded without a natural ecology of infant locomotion.
Researchers know nothing about how much infants crawl and
walk, how their activity is distributed over time, how far they
travel and where they go, how frequently they fall and what
motivates them to persevere, and how natural locomotion
changes with development. Lack of such descriptive data is a
serious omission, unique to motor development. Descriptions
of infants’ natural activity have been instrumental for constraining theory, guiding clinical interventions, and motivating
new research in other areas, such as language acquisition
(Hart & Risley, 1995; Hurtado, Marchman, & Fernald, 2008;
MacWhinney, 2000), cognitive development (Piaget, 1936/
1952), social-emotional development (Barker & Wright, 1951;
Messinger, Ruvolo, Ekas, & Fogel, 2010), symbolic play
(Tamis-LeMonda & Bornstein, 1996), sleep (Kleitman &
Engelmann, 1953), and natural vision (Cicchino, Aslin, &
Rakison, 2011; Franchak, Kretch, Soska, & Adolph, 2011;
Smith, Yu, & Pereira, 2011). But theories about the development of locomotion and therapies designed to redress atypical
locomotor development are not connected to data on infants’
real-world experiences with locomotion.
Why are natural descriptions so conspicuously absent from
the literature on infant locomotion? One reason for the absence
of data is the traditional emphasis on neuromuscular maturation. The long-held assumption that locomotion develops as a
universal series of increasingly erect stages led researchers to
focus on the formal structure of prone crawling postures en
route to upright walking (Gesell, 1946). Similarly, the search
for locomotor “primitives” led to formal comparisons between
alternating leg movements in newborn stepping, treadmillelicited stepping, and independent walking (Dominici et al.,
2011; Forssberg, 1985; McGraw, 1945; Thelen, 1986; Zelazo,
1983). But age-related sequences in the topography of locomotion dodge the question of why crawlers ever bother to
walk. That is, why would expert crawlers abandon a presumably stable, quadrupedal posture that took months to master in
order to move in a precarious, upright posture that involves
frequent falling? In fact, the question of why children persist
Corresponding Author:
Karen E. Adolph, Department of Psychology, New York University, 4
Washington Place, Room 410, New York, NY 10003
E-mail: [email protected]
1388
in acquiring new skills that are initially less functional than the
skills already in their repertoires is a central but unanswered
question in developmental psychology (Miller & Seier, 1994;
Siegler, 2000).
A second, related, reason for the lack of data on natural
locomotion is that researchers have historically measured
aspects of periodic gait—consecutive, regular steps over open
ground—rather than natural locomotion in a cluttered environment where deviations from periodic gait can be adaptive and
functional. Since the 1930s, researchers have described
infants’ movements as they take a series of continuous steps
over a straight, uniform path (Bril & Breniere, 1993; Dominici
et al., 2011; Hallemans, De Clercq, & Aerts, 2006; McGraw,
1945; Shirley, 1931). With the standard paradigm, it is imperative that the infants being assessed walk as quickly and as
straight as possible because speed and straightness affect measures such as step length. But the first thing that researchers
discover as they try to coax infants along a straight, continuous
path is that infants do not readily walk this way. Instead, they
stop after a few steps, speed up and slow down, swerve and
change direction, and misstep or fall. Typically, such deviations from periodic gait are ignored because they invalidate
standard skill measures, and thus trials must be repeated.
However, in natural locomotion, modifications in step length,
speed, and direction are necessary to cope with variable terrain
(Patla, 1997). Without a corpus of natural infant locomotion,
researchers cannot know whether standard skill measures such
as step length and speed during periodic gait are related to
functional skill measures in the everyday environment, such as
how much infants crawl or walk, how many steps they take,
how far they travel, and how frequently they fall.
A third factor contributing to ignorance about infants’ natural experiences with locomotion is that researchers (including
the current authors) routinely represent experience as the number of days that have elapsed since an onset date. Researchers
report walking experience as the number of days between the
first day of walking and the day of testing. However, this definition is misleading: New walkers walk intermittently, vacillating between days when they walk and days when they do
not (Adolph, Robinson, Young, & Gill-Alvarez, 2008). More
important, this definition is a conceptual misrepresentation of
experience. The passage of time is only a proxy for the events
that infants actually experience (Adolph & Robinson, in press;
Wohlwill, 1970). Although walking experience reliably predicts improvements in standard skill measures such as step
length and step width (Adolph, Vereijken, & Shrout, 2003;
Bril & Breniere, 1993) and in performance of perceptualmotor tasks such as perceiving affordances of slopes (Adolph,
1997), the number of days since walking onset carries little
more meaning than test age (the number of days since birth).
Indeed, some researchers refer to the number of days since
onset as “walking age” (Clark, Whitall, & Phillips, 1988). Possibly, sheer practice, indexed by accumulated number of steps,
facilitates improvements in gait. Alternatively, particular
experiences, such as surfaces encountered or falls, may teach
Adolph et al.
infants to walk. Without a natural corpus of infant locomotion,
there is no empirical basis for hypothesizing about underlying
learning mechanisms.
The Current Study
The current study provides the first data on natural infant locomotion—time in motion and distribution of activity over time,
variety of locomotor paths, and accumulated steps, distance,
and falls. We had three aims. First, we compared natural locomotion in experienced crawlers and novice walkers to gain
purchase on the question of why crawlers are motivated to
walk. Second, we asked whether functional measures of walking skill, such as number of steps and number of falls per hour,
improve with test age and walking age, as do standard skill
measures like step length and step width. Third, we investigated relations between standard and functional measures of
walking skill.
Presumably, most spontaneous walking occurs while
infants explore the environment and interact with caregivers.
Accordingly, data were collected while infants played freely
under caregivers’ supervision. We videotaped infants rather
than relying on step counters or parent informants—two methods that proved problematic in earlier attempts to quantify
infants’ natural locomotion (Adolph, 2002). Because video
coding was intensely detailed and laborious, we collected representative (15- to 60-min) samples of activity, as is customary
in studies of language acquisition (e.g., Hurtado et al., 2008).
Most samples were collected in a laboratory playroom to maximize recording quality and to eliminate individual differences
in infants’ home environments. We also observed infants in
their homes to ensure the validity of the laboratory data for
estimating natural activity. We focused on 12- to 14-month-old
novice walkers, in whom improvements in standard skill measures are most dramatic, but included a sample of older, more
experienced 19-month-olds, whose skill measures typically
have begun to reach asymptote (Adolph et al., 2003; Bril &
Breniere, 1993; Clark et al., 1988; Hallemans et al., 2006). We
also observed a comparison group of 12-month-old expert
crawlers.
Method
Participants and procedure
We collected 15 to 60 min of spontaneous activity for 151
infants (72 girls, 79 boys) from the New York City area. Most
families were middle-class, and 73% were White. Data from 5
additional infants were excluded because of fussiness or technical problems. We observed 20 crawlers (11.8–12.2 months of
age) and 116 walkers (11.8–19.3 months) in a laboratory playroom (8.66 m × 6.10 m) filled with furniture, varied ground
surfaces, and toys (Fig. 1a). Infants could move freely throughout the room (Fig. 1b). To ensure that playroom observations
were representative of natural locomotion, we also observed
1389
Walking Experience
a
a
Number of Infants
24
18
12
6
0
0
60
120
180
240
300
Walking Age (days)
b
Crawlers
n = 20
b
Age (months)
12
Walkers
n = 20
12
13
n = 30
13
n = 15
14
n = 36
19
n = 30
0
60
120
180
240
300
Crawling or Walking Age (days)
Fig. 2. Frequency histogram of walking age across the entire sample (a)
and distribution of crawling and walking ages for the six groups of infants
(b). In (a), striped portions of the bars denote girls, and nonstriped portions
denote boys. In both (a) and (b), gray bars denote home observations, and
white bars denote lab observations. Each vertical line in (b) represents
one infant.
Fig. 1. Layout of the laboratory playroom (a) and an example of a natural
walking path (b). Dimensions are drawn to scale. In (a), the large rectangle
on the left shows the location of the gait carpet and a representative
walking path over the carpet. The playroom also contained a couch, a
padded square pedestal, a slide and small stairs, a narrow catwalk behind
a wooden barrier, large steps at the ends of the catwalk, a set of carpeted
stairs, a set of wooden stairs, a standing activity table, and a wall lined with
shelves of toys. The line superimposed over the diagram in (b) shows the
natural walking path of one typical 13-month-old during the first 10 min of
spontaneous play. Overlapping lines indicate revisits to the same location.
Filled circles represent the location of rest periods longer than 5 s; open
circles denote falls.
fifteen 12.8- to 13.8-month-old walkers in their homes. Caregivers were instructed to interact normally with their infants
and to mind their safety. In both settings, an experimenter
recorded infants’ movements with a handheld camera. In the
laboratory, two additional fixed cameras recorded side and
overhead views to aid coding.
Crawling and walking age were determined from parental
reports of the first day that infants traveled 10 ft across a room
without stopping. Walking age was unavailable for 5 infants.
Figure 2a shows a frequency distribution of walking age and
sex, and whether infants were observed in the lab playroom or
in their homes. Figure 2b shows the distribution of crawling
and walking ages for the six groups of infants: twenty 12month-old crawlers (observed in the lab for 20 min), twenty
12-month-old walkers (observed in the lab for 20 min), thirty
13-month-old walkers (observed in the lab for 30 min), thirtysix 14-month-old walkers (observed in the lab for 15 min),
thirty 19-month-old walkers (observed in the lab for 30 min),
and fifteen 13-month-old walkers (observed at home for 60
min). Notably, in the 12-month-olds, crawling age (M = 97.6
days) was considerably larger than walking age (M = 29.7
days), t(38) = 5.41, p < .01 (see the top two rows of Fig. 2b).
Across the entire sample, walking age ranged from 5 to 289
days. Walking age overlapped among the 12- to 14-montholds, and there was no difference in walking age between
13-month-olds observed at home (M = 47.4 days) and in the
lab (M = 45.9 days), p > .10.
At the end of the laboratory sessions, we collected two
standard measures of walking skill as infants walked a straight
path over a pressure-sensitive mat (3.6 m × 0.89 m; GAITRite
System, CIR Systems, www.gaitrite.com; see Fig. 1a): step
length (front-to-back distance between consecutive footfalls)
1390
Adolph et al.
To characterize the overall amount of natural locomotion, we
calculated the accumulated time crawling or walking, number
of steps, and number of falls for each infant and then expressed
the data as proportions or hourly rates to allow comparisons
across groups that were observed for different amounts of time.
We estimated the total distance that infants walked, as if stringing their steps together end to end, by multiplying each infant’s
total step number by his or her average step length on the gait
carpet.
and step width (side-to-side distance between feet). We estimated crawlers’ average step length (distance between consecutive knee contacts) from the number of steps taken to
crawl a 3.6-m path. Three walkers and 1 crawler did not contribute usable data.
Data coding
A primary coder scored 100% of the video data for the duration of time crawling or walking, number of crawling or walking steps, and number of falls. Time crawling or walking was
the duration of a single step or series of steps flanked by rest
periods of at least 0.5 s; onsets were scored from the video
frame when the walker’s foot (or crawler’s knee) left the floor,
and offsets were scored from the video frame when the foot or
knee touched the floor in the last step of the series. Coders did
not score time in motion for the home observations because
they could not determine bout onsets and offsets reliably.
A step was considered any up-and-down motion of a leg that
changed the infant’s location on the floor. Falls were scored
when infants lost balance while crawling or walking, and their
bodies dropped to the floor unsupported. A second coder
independently scored 25% of each infant’s data. Interrater
agreement was high for time crawling or walking, number of
steps, and number of falls, rs > .95, ps < .01.
c
*
25
0
.75
.50
.25
2,000
1,000
.00
Crawler
Walker
e
f
n.s.
Crawler
Steps/Falls
200
100
0
g
n.s.
Walker
Crawler
Walker
n.s.
30
200
100
20
10
0
0
Crawler
250
Walker
300
300
Time (s)/Falls
Walker
500
0
0
Distance (m)/Falls
Crawler
*
750
Distance (m)/Hour
50
d
*
3,000
1.00
Time in Motion
75
Falls/Hour
b
*
How did functional skill measures compare between 12-monthold crawlers and walkers? As expected, novice walkers fell
more times per hour (M = 31.5) than expert crawlers did (M =
17.4), t(38) = 2.52, p < .02 (Fig. 3a), although the prevalence
of falls in expert crawlers was unexpected. However, walkers
walked more than crawlers crawled (Figs. 3b–3d): Walkers
spent a larger proportion of time in motion (M = .33) than
crawlers (M = .20), t(38) = 3.04, p < .01; walkers accumulated
more steps per hour (M = 1,456.1) than crawlers (M = 635.9),
t(38) = 3.78, p < .01; and walkers traveled greater distances
per hour (M = 296.9 m) than crawlers (M = 100.4 m), t(36) =
4.05, p < .01. When we reconsidered falls taking into account
the differences in activity between crawlers and walkers, differences in fall rate disappeared (Figs. 3e–3g): For every fall,
Steps/Hour
a
Results
Crawler
Walker
Crawler
Walker
Fig. 3. Comparisons between 12-month-old expert crawlers and 12-month-old novice walkers: (a) number of falls per hour, (b) proportion
of time in motion, (c) number of steps per hour, (d) distance traveled per hour, (e) accumulated time in motion for each fall, (f) accumulated
number of steps for each fall, and (g) accumulated distance traveled for each fall. Solid horizontal lines in these box plots denote medians, and
dashed horizontal lines denote means; circles denote outliers beyond the 10th and 90th percentiles; boxes include the 25th to 75th percentiles;
tails denote the 10th and 90th percentiles. Asterisks indicate significant differences between the two groups, p < .05.
1391
Walking Experience
Table 1. Correlations Between Test Age, Walking Age, and Skill Measures
Measure
Walking age
Step length
Step width
Time walking
Steps/hour
Distance/hour
Falls/hour
Test age
Walking age
Step length
Step width
Time walking
Steps/hour
Distance/hour
.86** (124)
.71** (111)
.74** (106)
−.60** (111)
−.68** (106)
−.65** (111)
.20* (114)
.28** (109)
.28** (111)
−.24* (111)
.46** (129)
.48** (124)
.51** (111)
−.42** (111)
.85** (114)
.65** (111)
.68** (106)
.76** (111)
−.55** (111)
.72** (111)
.92** (111)
−.35** (129)
−.33** (124)
−.28** (111)
.32** (111)
.14 (114)
−.09 (129)
−.17 (111)
Note: Degrees of freedom are shown in parentheses.
*p < .05. **p < .01.
walkers accumulated 1.2 min in motion, on average, and
crawlers accumulated 1.7 min; walkers accumulated 69.2
steps for each fall, on average, and crawlers accumulated 54.7;
walkers traveled 12.5 m for each fall, on average, and crawlers
traveled 8.6 m; all ps > .10.
Across the entire data set, walking infants averaged 2,367.6
steps per hour, traveled 701.2 m per hour, and fell 17.4 times
per hour. However, like periodic gait, natural walking develops (see Fig. S1 in the Supplemental Material available
online). As shown in the top two rows of Table 1, test age and
walking age were significantly correlated with both standard
measures of walking skill (step length, step width) and functional measures of walking skill (proportion of time walking,
number of steps per hour, distance traveled per hour, number
of falls per hour): Older infants (as identified by both chronological and walking age) took longer, narrower steps during
periodic gait over the gait carpet. And during free play, older
infants spontaneously spent more time walking, took more
steps, traveled farther distances, and fell less frequently, all
ps < .01. The significant correlations between age and functional skill measures remained even when time in motion was
partialed out (Table 2), all ps < .01, which means that functional skill measures reflect more than overall activity level.
Also, infants observed in their homes appeared similar to
infants observed in the laboratory playroom; t tests comparing
home (n = 15) and lab (n = 70) observations of infants with
equivalent walking age showed no differences in number of
Table 2. Partial Correlations Between Test Age, Walking Age,
and Skill Measures Controlling for Time Walking
Measure
Test age
Walking age
Step length
Step width
Steps/hour
Distance/hour
Steps/hour
Distance/hour
Falls/hour
.55** (113)
.49** (108)
.54** (110)
−.43** (110)
.75** (110)
.71** (105)
.84** (110)
−.57** (110)
.84** (110)
−.39** (113)
−.39** (109)
−.33** (110)
.36** (110)
−.39** (113)
−.40** (110)
Note: Degrees of freedom are shown in parentheses.
**p < .01.
steps per hour or number of falls per hour, ps > .10 (Figs. S1d
and S1e).
Infants who were better walkers on the gait carpet were also
better walkers during free play: Standard and functional skill
measures were significantly correlated (Table 1), and these
correlations remained after time in motion was partialed
out (Table 2). Time walking, number of steps per hour, and
distance traveled per hour were inherently intercorrelated
(Table 1) because infants who took more steps had to cover
more ground and spend more time in motion. However,
number of falls per hour was not correlated with time walking,
number of steps per hour, or distance traveled per hour
(Table 1) because although infants who walked more had more
opportunities to fall, they were also better walkers and thus fell
less. When time in motion was partialed out, number of falls
per hour was significantly negatively correlated with number
of steps per hour and distance traveled per hour (Table 2), and
all functional measures were consistent: Better walkers took
more steps, traveled farther distances, and fell less frequently.
Although standard and functional skill measures were correlated, periodic gait on the gait carpet and natural locomotion
during free play looked very different (Figs. 1a and 1b). Our
impression from scoring the video files was that infants’ natural paths twisted through most of the open space in the room.
We confirmed that impression in 7 randomly selected novice
walkers (mean walking age = 57.7 days) and 7 experienced
walkers (mean walking age = 190.3 days) in the first 10 min of
play. We superimposed 105 grid squares over the open areas of
the playroom and scored each time the infants entered each
square. All infants rambled throughout the room and spontaneously played near the couch and on the slide, pedestal,
catwalk, carpeted stairs, and wooden stairs. The number of different grid squares entered was similar between novices (M =
49) and experts (M = 57), but experts made more return trips
to the same squares. Novices entered or reentered 128.3 grid
squares, on average, and experts entered or reentered 205.9
grid squares, t(12) = 2.71, p < .05.
Although infants accumulated thousands of steps during
the observation periods, they spent most of the time stationary.
They were under no obligation to move, and one 12-monthold did not take any walking steps. On average, infants walked
1392
only 32.3% of the time. Walking was distributed over time in
primarily short bursts of activity. The raster plot in Figure S2
in the Supplemental Material shows the even distribution of
walking bouts for the 60 infants observed for 30 min, ranked
by walking age. Raster plots of the other 56 infants for whom
we scored bout duration showed similarly even distributions.
On average, 46% of bouts consisted of one to three steps, and
23% consisted of a single step—too short to qualify as periodic gait and too short for calculating standard measures of
walking skill. There was no difference in duration, step number, or step rate between walking bouts that ended in falls and
those that did not, ps > .10.
Discussion
A remarkable thing about basic skills acquired during infancy is
the apparent ease and rapidity of their acquisition. Infants learn
to walk, talk, think, play, and perceive objects and events in the
course of natural activity. Thus, descriptions of natural activity
play a critical role in guiding research, theory, and application.
The development of locomotion is a notable exception: Until
now, research, theory, and clinical intervention have proceeded
without a natural ecology of infant locomotion. By collecting
such a corpus, we aimed to address the question of why expert
crawlers transition to walking, investigate developmental
changes in natural locomotion and whether they relate to
improvements on standard measures, and provide an empirical
basis for hypothesizing about learning mechanisms.
Why walk?
Our inclusion of a comparison group of expert crawlers provided some clues to the long-standing puzzle of why infants
who are skilled crawlers abandon crawling for a precarious,
new, upright posture. To our surprise, expert crawlers were not
more skilled than novice walkers. Functional measures of
locomotor skill showed that crawlers crawled less than walkers walked, took fewer steps, and traveled shorter distances.
Moreover, falling was common: All but one crawler fell. As
expected, falling was far more common in novice walkers:
One racked up 69 falls per hour. But when we reconsidered
fall rate to take into account the differences in activity between
crawlers and walkers, the difference in fall rates disappeared,
and walkers were no longer at a disadvantage. In fact, when
we reanalyzed standard measures of locomotor skill (measures
of crawling or walking over a straight, uniform path) in infants
observed longitudinally (originally reported in Adolph, 1997),
step length and speed increased steadily from infants’ 1st week
of crawling to their 19th week of walking, and showed no decrement over the transition from crawling to walking (Adolph,
2008). In other words, assessments of both standard and functional skill measures indicate that new walkers reap all the
benefits of an upright posture without incurring additional
risk of falling. Thus, part of the answer to “why walk?” is
“why not?”
Adolph et al.
Development of natural locomotion
After 100 years of studying the development of walking by
coercing infants to walk at a steady pace along a straight, uniform path, researchers can say with certitude that standard
measures of walking skill (e.g., step length and step width)
improve with test age and walking age. We replicated that
century-old finding. More newsworthy is our finding that natural locomotion also improves: Functional measures of walking skill obtained from spontaneous locomotion during free
play (number of steps, distance traveled, and number of falls
per hour) improve with test age and walking age. These findings held up after statistically adjusting for time walking,
which means that older, more experienced walkers not only
walk more, but also walk better. Just as standard skill measures are intercorrelated, functional skill measures were highly
consistent. When time walking was partialed out to statistically adjust for activity, analyses showed that infants who took
more steps and traveled farther distances fell less frequently.
Moreover, we found that standard and functional skill measures were significantly correlated. Thus, this study provides
the first evidence of construct validity for standard skill measures in terms of natural infant walking. This set of findings is
remarkable because periodic gait (Fig. 1a) looks notably different from natural locomotion (Fig. 1b).
Possible learning mechanisms
Researchers need to reconsider the long-held tradition of using
walking age to represent walking experience. Walking age signifies only the elapsed time since walking onset. Like test age,
walking age is a robust predictor of various developmental
outcomes, but it is not an explanatory variable. In other areas
of developmental research, descriptions of natural activity
have informed understanding about learning mechanisms. For
example, in language acquisition, the sheer number of utterances and word tokens in mothers’ natural talk to infants when
they are 18 months old (estimated from 12 min of motherinfant free play in a laboratory playroom) predicts their rate of
vocabulary growth and language processing speed at 24
months of age (Hurtado et al., 2008). In contrast, diversity of
language (number of word types) is not predictive. In conceptual development, event type rather than sheer quantity of
input affects learning about causal agency: A higher proportion of agentive events compared with self-propulsion events
(estimated from 1 hr of video collected with a head camera)
during natural activity at 3, 8, and 12 months of age influences
generalization about causal agency (in habituation tasks) at 10
to 14 months of age (Cicchino et al., 2011). Similarly, a corpus
of natural locomotion allows researchers to investigate possible learning mechanisms by analyzing specific measures of
locomotor experience. The current study suggests that quantity, distribution, and variety of experiences are viable candidates as factors affecting learning to walk.
Although most people would assume that infants walk and
fall a lot, few would guess that the average toddler takes 2,368
1393
Walking Experience
steps, travels 701 m—the length of 7.7 American football
fields—and falls 17 times per hour. Hourly rates provide only
a tantalizing window into the amounts of practice that likely
accumulate over a day. For example, a multiplier of 6 hr
(approximately half of infants’ waking day) would indicate
that infants take 14,000 steps daily, travel the length of 46
football fields, and incur 100 falls. Estimates of natural activity are equally enormous for other skills. Middle-class infants
hear 2,150 words per hour, more than 30 million words by 3
years (Hart & Risley, 1995). Eleven- to 13-month-olds spend
more than 30 min per hour engaged with objects during everyday activity (Karasik, Tamis-LeMonda, & Adolph, 2011). By
2 months of age, infants have executed more than 2.5 million
eye movements (Johnson, Amso, & Slemmer, 2003), and by
3.5 months, they have performed 3 to 6 million.
To put these immense numbers into perspective, consider
that concert musicians and professional athletes require
approximately 4 hr of practice per day to train and fine-tune
their perceptual-motor systems (Ericsson, Krampe, & TeschRomer, 1993). The consensus in the literature on expertise is
that large amounts of regular practice, accumulated over years
of training, promote expert performance (Ericsson & Ward,
2007). The same principle could apply to acquiring expertise
in walking.
Natural walking was distributed in time and occurred in
variable patterns and contexts. Short bursts of walking were
separated by longer stationary periods. Walking bouts were
frequently too short—one to three steps—to qualify as periodic gait. Moreover, infants started and stopped at will, traveled in winding paths over varying surfaces, took sideways
and backward steps, varied their walking speed, switched
from upright to other postures, and misstepped and fell. They
visited multiple locations and engaged in different activities
therein.
Laboratory studies with older children and adults indicate
that time-distributed, variable practice is beneficial for motor
learning (Gentile, 2000; Schmidt & Lee, 1999). Time-distributed
practice is more effective than massed practice because
intermittent rest periods allow learning to be consolidated,
relieve fatigue, and renew motivation. Variable practice leads
to greater flexibility and broader transfer than blocked practice
because executing a variety of movements in a variety of contexts helps learners to identify the relevant parameters and
their allowable settings. Recent efforts to teach robots to walk
provide additional support for the effectiveness of variable
practice. The traditional approach is to train robots to walk as
fast as possible in a straight line—essentially, to train robots
on periodic gait (Kohl & Stone, 2004). But training robots
with omnidirectional gait on variable paths—a regimen similar to infants’ natural locomotion—led to more adaptive, functional locomotor skill. After 15,000 runs through an obstacle
course, robots had fewer falls, took more steps, traveled
greater distances, and moved more quickly than they had prior
to training. Moreover, in a test not possible with infants, they
exhibited elite performance in robot soccer: With a variable
training regimen, the UT Austin Villa team won all 24 games
in the 2011 RoboCup 3D simulation competition, scoring 136
goals and conceding none (MacAlpine, Barrett, Urieli, Vu, &
Stone, 2012; Urieli, MacAlpine, Kalyanakrishnan, Bentor, &
Stone, 2011).
Conclusion
How do infants learn to walk? This corpus of natural locomotion indicates that infants accumulate massive amounts of
time-distributed, variable practice. Over days of walking, they
take more steps, travel farther distances, and fall less. And
they may be motivated to walk in the first place because walking takes them farther faster than crawling without increasing
the risk of falling. Traditional studies of infant locomotion
during periodic gait could not have revealed these findings.
Acknowledgments
We thank Mark Blumberg, Judy DeLoache, Peter Gordon, Rachel
Keen, Patrick MacAlpine, and Peter Stone for helpful comments;
Samira Iravani for the line drawings in Figure 1; and John Franchak
for help with other figures.
Declaration of Conflicting Interests
The authors declared that they had no conflicts of interest with
respect to their authorship or the publication of this article.
Funding
This research was supported by National Institute of Child Health and
Human Development Grant R37-HD033486 to K. E. A. Additional
funding was provided by the Cattell Fund and Procter & Gamble.
Supplemental Material
Additional supporting information may be found at http://pss.sagepub
.com/content/by/supplemental-data
References
Adolph, K. E. (1997). Learning in the development of infant locomotion. Monographs of the Society for Research in Child Development, 62(3, Serial No. 251).
Adolph, K. E. (2002). Learning to keep balance. In R. Kail (Ed.),
Advances in child development and behavior (Vol. 30, pp. 1–40).
Amsterdam, The Netherlands: Elsevier Science.
Adolph, K. E. (2008). The growing body in action: What infant
locomotion tells us about perceptually guided action. In R. L.
Klatzky, B. MacWhinney, & M. Behrmann (Eds.), Embodiment,
ego-space, and action (pp. 275–321). New York, NY: Taylor &
Francis Group.
Adolph, K. E., & Robinson, S. R. (in press). The road to walking:
What learning to walk tells us about development. In P. Zelazo
(Ed.), Oxford handbook of developmental psychology. New York,
NY: Oxford University Press.
Adolph, K. E., Robinson, S. R., Young, J. W., & Gill-Alvarez, F.
(2008). What is the shape of developmental change? Psychological Review, 115, 527–543.
Adolph, K. E., Vereijken, B., & Shrout, P. E. (2003). What changes in
infant walking and why. Child Development, 74, 474–497.
1394
Barker, R. G., & Wright, H. F. (1951). One boy’s day: A specimen
record of behavior. New York, NY: Harper Brothers.
Bril, B., & Breniere, Y. (1993). Posture and independent locomotion
in early childhood: Learning to walk or learning dynamic postural control? In G. J. P. Savelsbergh (Ed.), The development of
coordination in infancy (pp. 337–358). Amsterdam, The Netherlands: Elsevier.
Cicchino, J. B., Aslin, R. N., & Rakison, D. H. (2011). Correspondences between what infants see and know about causal and selfpropelled motion. Cognition, 118, 171–192.
Clark, J. E., Whitall, J., & Phillips, S. J. (1988). Human interlimb
coordination: The first 6 months of independent walking. Developmental Psychobiology, 21, 445–456.
Dominici, N., Ivanenko, Y. P., Cappellini, G., D’Avella, A., Mondi,
V., Cicchese, M., & . . . Lacquaniti, F. (2011). Locomotor primitives in newborn babies and their development. Science, 334,
997–999.
Ericsson, K. A., Krampe, R. T., & Tesch-Romer, C. (1993). The role
of deliberate practice in the acquisition of expert performance.
Psychological Review, 100, 363–406.
Ericsson, K. A., & Ward, P. (2007). Capturing the naturally occurring
superior performance of experts in the laboratory: Toward a science of exceptional performance. Current Directions in Psychological Science, 16, 346–350.
Forssberg, H. (1985). Ontogeny of human locomotor control I. Infant
stepping, supported locomotion, and transition to independent
locomotion. Experimental Brain Research, 57, 480–493.
Franchak, J. M., Kretch, K. S., Soska, K. C., & Adolph, K. E. (2011).
Head-mounted eye tracking: A new method to describe infant
looking. Child Development, 82, 1738–1750.
Gentile, A. M. (2000). Skill acquisition: Action, movement, and neuromotor processes. In J. Carr & R. Shepard (Eds.), Movement
science: Foundations for physical therapy in rehabilitation (2nd
ed., pp. 111–187). New York, NY: Aspen Press.
Gesell, A. (1946). The ontogenesis of infant behavior. In L. Carmichael
(Ed.), Manual of child psychology (pp. 295–331). New York, NY:
John Wiley.
Hallemans, A., De Clercq, D., & Aerts, P. (2006). Changes in 3D
joint dynamics during the first 5 months after the onset of independent walking: A longitudinal follow-up study. Gait & Posture,
24, 270–279.
Hart, B., & Risley, T. D. (1995). Meaningful differences in the everyday experiences of young American children. Baltimore, MD:
Paul H. Brookes.
Hurtado, N., Marchman, V. A., & Fernald, A. (2008). Does input
influence uptake? Links between maternal talk, processing speed
and vocabulary size in Spanish-learning children. Developmental
Science, 11, F31–F39.
Johnson, S. P., Amso, D., & Slemmer, J. A. (2003). Development of
object concepts in infancy: Evidence for early learning in an eye
tracking paradigm. Proceedings of the National Academy of Sciences, USA, 100, 10568–10573.
Karasik, L. B., Tamis-LeMonda, C. S., & Adolph, K. E. (2011).
The transition from crawling to walking and infants’ actions with
objects and people. Child Development, 82, 1199–1209.
Adolph et al.
Kleitman, N., & Engelmann, T. G. (1953). Sleep characteristics of
infants. Journal of Applied Physiology, 6, 269–282.
Kohl, N., & Stone, P. (2004). Policy gradient reinforcement learning for fast quadrupedal locomotion. Proceedings of the IEEE
International Conference on Robotics and Automation, 3, 2619–
2624.
MacAlpine, P., Barrett, S., Urieli, D., Vu, V., & Stone, P. (2012).
Design and optimization of an omnidirectional humanoid walk:
A winning approach at the RoboCup 2011 3D simulation competition. In Proceedings of the Twenty-Sixth AAAI Conference on
Artificial Intelligence (AAAI-12) (Vol. 1, pp. 1047–1053). Palo
Alto, CA: Association for the Advancement of Artificial Intelligence.
MacWhinney, B. (2000). The CHILDES project: Tools for analyzing
talk. Mahwah, NJ: Erlbaum.
McGraw, M. B. (1945). The neuromuscular maturation of the human
infant. New York, NY: Columbia University Press.
Messinger, D., Ruvolo, P., Ekas, N., & Fogel, A. (2010). Applying
machine learning to infant interaction: The development is in the
details. Neural Networks, 23, 1004–1016.
Miller, P. H., & Seier, W. (1994). Strategy utilization deficiencies in
children: When, where, and why. In H. Reese (Ed.), Advances
in child development and behavior (Vol. 25, pp. 107–156). New
York, NY: Academic Press.
Patla, A. E. (1997). Understanding the role of vision in the control of
human locomotion. Gait & Posture, 5, 54–69.
Piaget, J. (1952). The origins of intelligence in children (M. Cook,
Trans.). New York, NY: International Universities Press. (Original work published 1936)
Schmidt, R. A., & Lee, T. D. (1999). Motor control and learning: A
behavioral emphasis (3rd ed.). Champaign, IL: Human Kinetics.
Shirley, M. M. (1931). The first two years: A study of twenty-five
babies. Westport, CT: Greenwood Press.
Siegler, R. S. (2000). The rebirth of children’s learning. Child Development, 71, 26–35.
Smith, L. B., Yu, C., & Pereira, A. F. (2011). Not your mother’s view:
The dynamics of toddler visual experience. Developmental Science, 14, 9–17.
Tamis-LeMonda, C. S., & Bornstein, M. H. (1996). Variation in children’s exploratory, nonsymbolic, and symbolic play: An explanatory multidimensional framework. In C. K. Rovee-Collier & L. R.
Lipsitt (Eds.), Advances in infancy research (Vol. 10, pp. 37–78).
Westport, CT: Ablex.
Thelen, E. (1986). Treadmill-elicited stepping in seven-month-old
infants. Child Development, 57, 1498–1506.
Urieli, D., MacAlpine, P., Kalyanakrishnan, S., Bentor, Y., & Stone,
P. (2011). On optimizing interdependent skills: A case study in
simulated 3D humanoid robot soccer. In Proceedings of the 10th
International Conference on Autonomous Agents and Multiagent
Systems (Vol. 2, pp. 769–776). Richland, SC: International Foundation for Autonomous Agents and Multiagent Systems.
Wohlwill, J. P. (1970). The age variable in psychological research.
Psychological Review, 77, 49–64.
Zelazo, P. R. (1983). The development of walking: New findings on
old assumptions. Journal of Motor Behavior, 2, 99–137.
Step Width (cm)
40
30
20
10
C 1.00
Prop. Time Walking
B
50
0
180
240
300
D
0.50
0.25
22
18
14
10
6
0
60
120
180
240
300
0
60
120
180
240
300
0
60
120
180
240
300
6000
4500
3000
1500
0
0.00
60
120
180
240
300
F
80
Distance / Hour (m)
Falls / Hour
120
0.75
0
E
60
Steps / Hour
Step Length (cm)
A
60
40
20
0
0
60
120
180
240
Walking Age (days)
300
2500
2000
1500
1000
500
0
Walking Age (days)
Scatterplots of walking age by (A-B) standard measures of walking skill obtained during
periodic gait over the gait carpet and (C-F) functional measures of walking skill obtained during
natural locomotion during free play. Circles denote infants observed in the laboratory playroom,
crosses denote infants observed at home.
Infants
0
5
10
15
20
Time (min)
25
30
Raster plot of distribution of walking bouts in 60 infants observed
over 30 minutes of natural walking. Each row represents one infant.
Rows were ordered by walking age. The width of each vertical line
represents the duration of the walking bout. The thinnest lines
denote bouts that were ≤ 6 s in duration (minimum bout duration
that could be represented given the resolution of the image).